Countries response to COVID19 is linked to individualism

rm(list = ls())
suppressPackageStartupMessages(library(ggplot2))
suppressPackageStartupMessages(library(tidyverse))
suppressPackageStartupMessages(library(plyr))
suppressPackageStartupMessages(library(dplyr))
suppressPackageStartupMessages(library(knitr))
suppressPackageStartupMessages(library(Rfast))
suppressPackageStartupMessages(library(ggpubr))
suppressPackageStartupMessages(library(jtools))
suppressPackageStartupMessages(library(httr))
suppressPackageStartupMessages(library(gganimate))
suppressPackageStartupMessages(library(transformr))
suppressPackageStartupMessages(library(broom.mixed))

Import data

Country population data from : https://datahub.io/JohnSnowLabs/population-figures-by-country#data

Country GDP data from: https://datahub.io/core/gdp#resource-gdp.

Covid confirmed cases data from John Hopkins CSSE

countries = read.csv('COUNTRY_AVERAGES_0406_rp.csv', stringsAsFactors = FALSE)
covid = read.csv("https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv", header=TRUE, stringsAsFactors = FALSE)[, -c(1,3,4)]

covid = covid %>% 
    group_by(Country.Region) %>%
    summarise_each(funs(sum)) %>% # sum to get one value for each country
    mutate(Country.Region=replace(Country.Region, Country.Region=="US", "United States"))

df = merge(countries, covid, by.x = "CountryofLiving", by.y = "Country.Region", all.x = TRUE)
df$lg_population = log10(df$population_2016)
df$lg_gdp = log10(df$gdp_2016)
df$lg_gdp_perperson = log10(df$gdp_2016/df$population_2016)
df$lg_percentConfirmed_320 = log10(df$X3.20.20/df$population_2016)
df$lg_percentConfirmed_latest = log10(df$X4.19.20/df$population_2016)

1. Association between individualism and covid behaviors

plot_asso = function(x, y){
  g = ggscatter(df, x = x, y = y,
          add = "reg.line",  
          conf.int = TRUE, 
          label = "CountryofLiving",
          repel = TRUE,
          add.params = list(color = "blue",
                            fill = "lightgray"))
  return(g)
}

Social distancing

plot_asso("individualism", "m_beh_distance")+
  stat_cor(method = "pearson", label.x = 55, label.y = 35) 

Hand washing

plot_asso("individualism", "m_beh_handwash")+
  stat_cor(method = "pearson", label.x = 20, label.y = 65) 

Social gathering

plot_asso("individualism", "m_beh_socgathering")+
  stat_cor(method = "pearson", label.x = 60, label.y = 76) 

Stay home

plot_asso("individualism", "m_beh_stayhome")+
  stat_cor(method = "pearson", label.x = 60, label.y = 100) 

Tell syptom

plot_asso("individualism", "m_beh_tellsymp")+
  stat_cor(method = "pearson", label.x = 60, label.y = 80) 

2. Association between COVID-severity and covid behaviors

Social distancing

plot_asso("lg_percentConfirmed_320", "m_beh_distance")+
  stat_cor(method = "pearson", label.x = -4, label.y = 50) +
  xlab("% confirmed cases in a population") + ylab("Social distancing")

Hand washing

plot_asso("lg_percentConfirmed_320", "m_beh_handwash")+
  stat_cor(method = "pearson", label.x = -5, label.y = 65) 

Social gathering

plot_asso("lg_percentConfirmed_320", "m_beh_socgathering")+
  stat_cor(method = "pearson", label.x = -4.5, label.y = 55) 

Stay home

plot_asso("lg_percentConfirmed_320", "m_beh_stayhome")+
  stat_cor(method = "pearson", label.x = -4.5, label.y = 55) 

Tell syptom

plot_asso("lg_percentConfirmed_320", "m_beh_tellsymp")+
  stat_cor(method = "pearson", label.x = -5, label.y = 75) 

3. Individualism is a strong predictor for %confirmed cases

Here we use %confirmed cases (\(log \frac{confirmed cases}{total population}\))as a measure to characterize the general severity of COVID-19 in a given country.

March 20th (Survey start date)

plot_asso("individualism", "lg_percentConfirmed_320")+
  stat_cor(method = "pearson", label.x = 55, label.y = -7) 

We find that after controlling for percent confirmed cases, individualism does not significantly predict social distancing behavior.

Comparison of models predicting social distancing behavior:

md1 = lm(m_beh_distance ~ individualism , data = df)
md2 = lm(m_beh_distance ~ individualism + lg_percentConfirmed_320, data = df)
md3 = lm(m_beh_distance ~ individualism + lg_gdp_perperson, data = df)

export_summs(md1, md2, md3,scale = FALSE,model.info = FALSE,model.fit = TRUE, model.names = c("model 1", "model 2", "model 3"), 
             coefs = c("individualism" = "individualism",
                      "%confirmed cases (log)" = "lg_percentConfirmed_320",
                      "GDP per person (log)" = "lg_gdp_perperson"))

	model 1	model 2	model 3
individualism	0.15 *	0.08	0.16
	(0.07)	(0.08)	(0.10)
%confirmed cases (log)		3.89
		(2.01)
GDP per person (log)			-0.97
			(4.66)
N	57	52	53
R2	0.09	0.15	0.08
*** p < 0.001; ** p < 0.01; * p < 0.05.

Today (2020-04-20)

plot_asso("individualism", "lg_percentConfirmed_latest")+
  stat_cor(method = "pearson", label.x = 70, label.y = -5) 

As a function of dates

df_long <- gather(df, year, confirmed_cases, X1.22.20:X4.19.20, factor_key=TRUE)

df_long$lg_percentConfirmed = log10(df_long$confirmed_cases/df_long$population_2016)

df_long$lg_percentConfirmed[!is.finite(df_long$lg_percentConfirmed)] <- NA
df_long2 = na.omit(df_long[,c(12,20,23, 25)])

df_long2$year = str_remove(df_long2$year, "[X]")
df_long2$date = as.Date(df_long2$year, format = "%m.%d.%Y")
dates = c()
beta = c()
ps = c()
for (date in unique(df_long2$date)){
  dates = c(dates, date)
  df_temp = df_long2[which(df_long2$date == date), ]
  correlation = cor.test(df_temp$individualism, df_temp$lg_percentConfirmed)
  md = lm(lg_percentConfirmed ~ individualism + lg_gdp_perperson, data = df_temp)
  s = summary(md)
  beta = c(beta, s$coefficients[2,1])
  ps = c(ps, s$coefficients[2,4])
}

df_dates = data.frame(dates, beta, ps)
df_dates$sig = "Not sig"
df_dates$sig[which(df_dates$ps < .05)] = "Sig"
keeps = seq(from = 1,to = 89, by = 5)
g = ggdotchart(df_dates, x = "dates", y = "beta",
           color = "sig",                                # Color by groups
           palette = c("#00AFBB", "#E7B800"), # Custom color palette
           add = "segments",                             # Add segments from y = 0 to dots
           ggtheme = theme_pubr())

ggpar(g, font.xtickslab = 7)

4. GDP (economic activities) is a strong predictor for %confirmed cases

March 20th (Survey start date)

plot_asso("lg_gdp_perperson", "lg_percentConfirmed_320")+
  stat_cor(method = "pearson", label.x = 4.5, label.y = -6.5) 

Today (2020-04-20)

plot_asso("lg_gdp_perperson", "lg_percentConfirmed_latest")+
  stat_cor(method = "pearson", label.x = 4.5, label.y = -6) 

5. Individualism does not predicts % confirmed cases after controlling for behavior intentions and GDP.

md5 = lm(lg_percentConfirmed_latest ~ individualism + m_beh_distance + lg_gdp_perperson, data = df)

summ(md5, confint = TRUE, vifs = TRUE)

Observations	52 (37 missing obs. deleted)
Dependent variable	lg_percentConfirmed_latest
Type	OLS linear regression

F(3,48)	43.75
R²	0.73
Adj. R²	0.72

	Est.	2.5%	97.5%	t val.	p	VIF
(Intercept)	-10.23	-11.59	-8.88	-15.15	0.00	NA
individualism	-0.00	-0.01	0.00	-1.38	0.17	1.87
m_beh_distance	0.01	0.00	0.02	2.81	0.01	1.09
lg_gdp_perperson	1.42	1.10	1.74	8.91	0.00	1.76
Standard errors: OLS